In a recent HFS Unfiltered conversation, Prashant Dinkar Hinge, CIO and Transformation Officer at MSIG USA, sat down with Saurabh Gupta and Ian Maher to discuss a sobering reality:
despite the noise, only about 5% to 7% of AI initiatives are actually scaling into production-grade environments.
If we want to be in that 7%, we need to stop treating AI as a “project” and start treating it as a marathon that requires specific training. Here is a peer-to-peer look at how we can move the needle.
As CIOs and business leaders in the insurance sector, you have all seen the cycle before. Every three to four years, the industry falls in love with a new acronym. Whether it was RPA, “Digital Transformation,” or Big Data, the hype cycle is a familiar old friend.
Today, that friend is Generative AI.
1. The Value Filter: Is it “P&L Real” or just “Marketing Real”?
Prashant makes a sharp distinction that IT leaders all need to bring to our boardrooms: AI remains a buzzword until it solves a business problem that is meaningful to the top or bottom line.
When evaluating partners or internal pilots, look past the “AI Practice” headcount and ask one question: “How much actual ‘lift’ are existing customers seeing?”. If the implementation isn’t moving hygiene metrics (speed and quality) or creating exponential differentiation for the customer, it’s just a marketing slogan.
2. The “Employee Triathlete” Model
One of the most profound shifts Prashant discusses is the evolution of talent. Do you really need 40-person implementation teams to deploy new solutions. Instead, small, agile teams of 3 to 5 “Triathletes” who possess three specific traits can do the same task:
- Business Context: They must understand the insurance value cycle—how a policy actually flows from underwriting to a claim.
- AI Literacy: They need to be fluent in “flow engineering” and prompting.
- Functional Expertise: They must remain masters of their specific domain, whether that’s actuarial science or IT infrastructure.
As leaders, your job isn’t just to buy the tools; it’s to win the “hearts and minds” of people so they bring their personal-life AI curiosity into their corporate roles.
3. Don’t Show Up to the Marathon Without Training
Prashant’s marathon analogy is a perfect warning for the IT leaders . You can’t just show up on “race day” (the day you decide to scale) and expect to finish if you haven’t done the prep work.
Scaling AI requires a foundation that many of the organizations are still building:
- Process Integration: AI sitting on top of siloed, broken processes will only “destroy value” elsewhere in the enterprise.
- Data Hygiene: AI is a powerful engine, but it won’t run on the “dirty fuel” of fragmented legacy data.
- Leadership Alignment: This is not a technology issue; it is a leadership issue. If the CEO and the direct reports aren’t aligned on the tactical methods of achieving productivity, the initiative will stall.
4. A Note to Sourcing Leadership: The Trust Dividend
For those of us managing vendor relationships, Ian Maher adds a critical perspective: Stop trying to force unilateral economic changes. Breaking a contract to demand “AI-driven cost savings” that haven’t been proven yet is a recipe for failure.
In this era, providers are learning as fast as enterprises are. You need to choose “Co-Learning” partners—those you trust to experiment with, share the risks, and ultimately share the rewards.
The Bottom Line
In the specialty and commercial space, the “moment of truth” remains the claims experience. If your AI investments aren’t making that moment better, faster, and more accurate, we are just adding to the noise.
Stop the “death by a thousand POCs” and start focused training for the marathon ahead.
Watch the full insight here: AI in Insurance: Buzzword or Real Transformation years, insurance finds a new obsession.
